Mechanistic-based Ductility Prediction for Complex Mg Castings

X SUN (PI) PACIFIC NORTHWEST NATIONAL LABORATORY

RICHLAND, WA, USA

2014 DOE VEHICLE TECHNOLOGY PROGRAM REVIEW JUNE 16-20, 2014

Project ID#: LM057

This presentation does not contain any proprietary, confidential, or otherwise restricted information

Overview

Timeline Start: Oct. 2010 End: Sep. 2014 80% Complete

Budget DOE - $1,800K

FY11 - $600k FY12 - $600k FY13 - $500k FY14 - $100k

Industries (in-kind) - $900K Industry - $300k/YR FY11-13

Barriers Limited ductility of Mg castings hindering its wider applications as vehicle components High ductility variations Lack of capability of conventional computational software/models in predicting ductility of Mg castings, resulting from various types of defects

Partners Ford Motor Company University of Michigan Mag-Tec Casting Corporation CANMET Materials Technology Laboratory

Project Objectives

Background and motivation Conventional computational technique (i.e., homogenization, continuum damage mechanics, crystal plasticity) and some phenomenological approaches have no or very limited ductility predictive capability for Mg castings

Objectives: To provide a modeling framework that can be used in future Mg alloy design and casting process optimization by

Developing an empirical casting process simulation tool that can estimate the variation in ductility and be used by the casting industry in the near future Developing a mechanistic-based predictive capability on key factors controlling Mg ductility that can be coupled with future advances in casting process simulation and will lead to further casting process optimization and alloy design

Deliverables

A validated simulation tool (quality map approach) for estimating the spatial variation of ductility and the influence of casting process variables (completed) Modeling and experimental methods in quantifying location-dependent intrinsic and extrinsic ductility limiting factors for complex Mg castings (completed 9/30/2013) Experimentally validated predictive models for stress versus strain curves, including ductility, for Mg castings considering both intrinsic and extrinsic ductility limiting factors (on-going; due 9/30/2014)

x

y

Technical Approaches

Cast a number of AM50/AM60 castings of complex geometries under a variety of conditions (i.e., melt temperature, shot speed, die temperature, gating geometry) – Ford, MagTec Ind., CANMET Mat. Tech. Lab. Perform alloying and casting process simulation to predict spatial variations in casting defects and other microstructural features under different conditions – Ford Characterize microstructure and defect features at various locations of the castings and perform tensile tests with samples machined from various locations – U. of Michigan, Ford Develop a quality-mapping capability for estimating/controlling ductility of Mg castings based on tensile test results and various casting parameters – U. of Michigan, Ford Develop a mechanistic-based ductility prediction capability with separate consideration of intrinsic factors and extrinsic factors - PNNL

Generic Closure Panel (GCIP)

Previous Ford Accomplishment

• Measured elongation with error versus predicted elongation. Agreement was acceptable except for 2 locations (circled)

• Need further understanding of statistical variation of elongation

Sample locations

Designed and manufactured new die cast insert to produce tensile and bend specimens

FY14 Accomplishments - Ford

FY14 Accomplishments - Ford

Test specimens to be used for determination of:

Statistical variability from machined versus as-cast samples Skin effects

FY14 Accomplishments - Ford

Retro-fit existing SEM to accommodate “in-situ” bending of Mg alloy samples Crack Propagation in samples with different aluminum content will be investigated to determine influence of:

β phase α phase Porosity skin

UM + Ford – Cast and Characterize Shrinkage Porosity and β-Mg17Al12 in SVDC AM Series

SEM 2.5 mm

SEM 5 mm

Scheil prediction

Average β area fraction increases as Al content is increased and as thickness is decreased. It is much lower than that predicted from “Scheil” solidification calculations

UM -- Developed quantitative methods for characterizing β-phase volume fraction and morphology

2.5mm

5mm

- Morphology changes from “isolated” at plate surface to “interconnected” in center of SVDC cast plates - β phase formation varies strongly through thickness of the plates and depends on plate thickness

0

1

2

3

4

5

6

7

8

9

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Area

Fra

ctio

n β

(%)

Normalized Plate Thickness

UM - Tested over 140 Tensile Samples from 16 SVDC Plates Produced by Ford

Model fit with

2.5mm

5mm

• Established major, statistically significant reference database quantifying the influence of composition and plate thickness on tensile behavior

• Quantified influence of Al composition and plate thickness on strain hardening exponent decreases

PNNL – Validated Intrinsic Model Considering Various β Volume Fraction and Morphology

α phase properties extracted from nanoindentation1,2

β phase properties calculated from first-principles3

Synthetic geometries generated to represent UM/Ford observations AM70

E = 38 GPa

5% σy = 130 MPa

5%

h’ = 0.1 MPa

10% ν = 0.35

AM40 E = 35 GPa

5%

σy = 120 MPa

5% h’ = 0.1 MPa

10%

ν = 0.35 β

AM70: 21% VF (t=10%) AM40: 8% VF (t=5%) E = 77.7 GPa ν = 0.35

May 3, 2014 14

1 Han et al., Materials Science and Engineering A, 473 (2008) 16-27 2 Shan, Gokhale, Materials Science and Engineering A, 361 (2003) 267-274 3 J. Wang, et al., Calphad, 35 (2011) 562-573

AM70 AM40

PNNL -- Effects of β Phase Morphology

May 3, 2014 15

Using AM40 Properties

Using AM70 Properties

8% β, t=4% 8% β, t=10% 8% β, t=30%

21% β, t=11% 21% β, t=30% 21% β, t=50%

Barker, et al., in press Comp. Mat. Sci., 2014.

PNNL -- Incorporating Intrinsic Microstructure-based Results into Extrinsic Models to Consider Pores

May 3, 2014 16

x

y

x

y

VF: 1% PS: avg 20µm (10~30µm)

Sun, et al., Mat. Sci. Eng. A, 2013. AM50 sample; Chadha, et al., Metal Mat Trans A, 2007

17

PNNL (Extrinsic Modeling) – Effects of Intrinsic Properties on Ductility

3D models with different properties for different layers

No.1

No.2 or No.4

No.3 or No.5

No.5 No.1 No.4

No.3 No.1

No.2

±10%

±25%

±10%

±25%

Input s-e curves for model #1,2,3 Resulted s-e curves

Input s-e curves for model #1,4,5 Resulted s-e curves The intrinsic property in the high-porosity core region has dominant effects on the ductility. The intrinsic property in the core may be used for 2D simulation.

18

PNNL (Extrinsic Factor Modeling) – Predicting Ductility based on Actual Microstructures with 2D/3D Correlation

*Song, Xiong, Li, Allison (2009) J. Alloys and Compounds, 477, pp.863-869.

Actual pore distributions *

3D simulation 2D simulation

Example of 3D/2D simulations of actual microstructure-based models

Ductility(%) Error (%)

(3D-Est.2D)/3D 2D model

3D model

Estimation from 2D

No. 1 3.40 5.18 4.60 11.16

No. 2 2.72 3.64 3.92 -7.55

No. 3 3.88 5.35 5.08 4.89

No. 4 4.67 5.85 5.88 -0.46

Element size :~25um εcr,tension : 20% Flatness : ~0.6

Comparison of 2D/3D ductility

Address Previous Year Reviewers’ Comments

Approach Reviewer 2: This reviewer pointed out that a mechanistic model would be a dramatic improvement; the reviewer was unclear how the necessary input parameters will be determined. The reviewer stated that overall the project draws heavily from experimentation, but the reviewer did not see how the modeling work will be fed back to affect the experimental work.

Response: The experimental work at different scales are designed to provide modeling parameters and serve as model validations. The initial modeling results on effects of b phase morphology and pore distribution have also guided the experimental design of solution treatment and HIP.

Reviewer 3: This reviewer opined that the work at Pacific Northwest National Laboratory (PNNL) always seems to be brute-force modeling of a particular system from which no generalizable insights can be drawn.

Response: The objective of this project is to develop a mechanism-based modeling framework that can be used to predict cast Mg ductility. We start with some specific AM alloys due to data availability and the general methodology should apply to other Mg systems.

Collaboration Reviewer 2: This reviewer observed that other participants do not seem as deeply involved as personnel at PNNL. Reviewer 3: This reviewer stated that not many details were provided other than Ford cast the parts, and Michigan modeled this.

Response: This is a PNNL-led project with active Ford and UM contribution as shown in the project structure and overall progress report. Ford has made significant contribution with the quality mapping for ductility and all microstructural and sample level characterizations are carried out by UM.

Future Work Reviewer 2: This reviewer pointed out that the remaining goals seem very broad and appear to be a large challenge to complete.

Response: We agree. We are currently on carryover from FY13 funding.

Collaborations

Ford, Mag-Tec Casting Corporation (Industry) Provided/operated high pressure casting and super vacuum die cast equipment Characterized coupon level stress versus strain curves for different conditions and locations Produced casting samples with varying aluminum content and performed casting process simulations Collaborated on characterization of microstructure and defect features at various locations on castings Developed and validated Mg casting quality map with statistical variation

University of Michigan (Academic) Established effects of plate thickness and aluminum contents on yield strength, ductility and UTS. Developed characterization techniques to quantify b volume fraction and morphology across sample thickness and for different aluminum contents Examined effects of solution treatment and hot isostatic pressing on hardening parameters and ductility

CANMET Materials Technology Laboratory Performed phase field simulation on pore and dendrite interactions

Summary Relevance

Limited ductility and high ductility variations hinder the wider applications of Mg casting in vehicle structures for lightweighting

Approach Develop a quality-mapping capability for estimating/controlling ductility of Mg castings based on tensile test results and various casting parameters Develop a mechanistic-based ductility prediction capability with separate consideration of intrinsic factors and extrinsic factors

Technical Accomplishments Validated quality mapping approach in predicting ductility and ductility variations Designed and manufactured new die cast insert for tensile and bending samples for in-situ SEM observation on α and β deformation mechanisms Quantified β-phase volume fraction and morphology for various aluminum content and casting plate thickness Validated intrinsic model considering various β volume fraction and morphology Predicted ductility based on actual pore structure produced by 3-D x-ray tomography

Collaborations Ford, Mag Tech, U Mich

Future work

Proposed Future Work Setting up in-situ SEM capability to examine crack growth under 3-point bend loading for alloys with different aluminum contents (Ford+UM) Use quantitative fractography and microstructural analysis to establish relationships between properties, microstructure and alloying/processing variables (UM+PNNL) Perform nano-indentation tests to evaluate grain boundary cohesive strength (PNNL/UM) Predict intrinsic ductility for Mg with different aluminum content by considering eutectic β phase and grain boundary decohesion (PNNL) Final deliverable – Validate prediction framework by comparing measured ductility with predicted ductility from x-ray tomography (PNNL)